CN103411970A - Alternating current transmission line insulator contamination condition detection method based on infrared thermography - Google Patents

Abstract

The present invention discloses an alternating current transmission line insulator contamination condition detection method based on infrared thermography, wherein an infrared thermal imager is adopted to shoot an insulator thermal field image, an image treatment technology is adopted to extract an insulator image, relevant characteristic quantities are calculated, an environmental temperature and an environmental humidity are combined, and genetically-optimized neural network training is adopted so as to judge a contamination level of the insulator in the infrared image. The detection method has the following advantages that: an insulator contamination condition can be detected under a condition of no power failure, installation of any equipment on a pole tower is not required, operation cost and maintenance cost are low, operations are simple, use is safe and reliable, and detection accuracy is high.

Description

Transmission line of alternation current insulator contamination condition detection method based on infrared thermal imagery

Technical field

The present invention relates to power transmission and transforming equipment running status maintenance field, relate in particular to a kind of transmission line of alternation current based on infrared thermal imagery insulator contamination condition detection method.

Background technology

Transmission line of electricity is in operational process, and the various particulates such as airborne dust, saline and alkaline, industrial fumes or birds droppings all can be deposited in outer surface of insulator and form pollution layer, make the insulator dielectric strength decreased, and pollution flashover easily occurs, and cause very large economic loss.If can effectively measure filthy degree and character, just likely avoid the power outage caused by pollution flashover.

In order to determine accurately cleaning or the flushing cycle of insulator, the method for the detection insulator contamination value adopted both at home and abroad at present mainly contains equivalent salt density method and leakage current method.At first the equivalent salt density method must have a power failure by after washing, measuring salinity density to transmission line of electricity, is difficult to the truth that reflects that insulator is in operation.The leakage current method is by detecting the size variation that flows through insulator surface pollution layer electric current under the working voltage effect, to detect the pollution level of insulator.Although the leakage current method can reflect more serious insulation fault, but after judging failure of insulation, leave the limited time that operating personnel process for, be difficult to independent widespread use, and it requires on each insulator chain to install a set of pick-up unit, high cost, the Maintenance and Repair of device need have a power failure and carry out.

Though infrared imaging method detects for insulator, there is no concrete insulator contamination deciding degree method, does not still obtain practical application in filthy context of detection.

Summary of the invention

Technical matters to be solved by this invention is that a kind of transmission line of alternation current based on infrared thermal imagery insulator contamination condition detection method will be provided, can fast and effeciently to the filthy state of current operation insulator, make accurately and reliably and analyzing, for clearing up in time insulator contamination, provide basis, to reduce the insulator contamination power outage.

Principle of the present invention is: when the shelf depreciation leakage current caused when pollution severity of insulators flows through megohmite insulant, dielectric loss or ohmic loss all can cause that the insulator local temperature raises, so take insulator thermal field image by thermal infrared imager, adopt image processing techniques to extract the insulator target area, calculate the correlated characteristic amount, combining environmental temperature and humidity, can judge the insulator contamination degree after the neural metwork training of genetic optimization.

The invention provides a kind of transmission line of alternation current based on infrared thermal imagery insulator contamination condition detection method, comprise the following steps:

(1) Infrared Thermogram of detected insulator picked-up;

(2) adopt neighborhood averaging to carry out the noise elimination to colored thermography;

(3) extract red R component and the green G component of thermography rgb space, compared to gray-scale map, G divides in spirogram the discrimination of insulator and background larger, and R divides insulator in spirogram to maintain higher pixel value, further auxiliary partition;

(4) adopt the seed region growth method to R, to divide spirogram and G to divide spirogram to cut apart respectively;

(5) R after cutting apart divides spirogram and G to divide spirogram to merge, and obtains final insulator and cuts apart image;

(6) extract the insulator contamination feature, characteristic quantity is: the maximum temperature in insulator chain card zone, minimum temperature, medial temperature, Temperature Distribution variance, three of the Temperature Distribution conic fitting parameters on the insulator axis;

(7) based on the GA-BP neural network, detect insulator dirty degree, take Insulator Infrared Image in service as sample graph, using environment temperature, humidity and seven characteristic quantities as input quantity, five gradation for surface pollution are as output quantity, set up a BP neural network, and utilizing genetic algorithm to be optimized the network node weights, the infrared image be written into temperature data namely obtains the gradation for surface pollution that this insulator is corresponding.

Superior effect of the present invention is:

1) Insulator Infrared Image of taking is carried out to denoising, improve picture quality and signal to noise ratio (S/N ratio), and according to the infrared image characteristics, choose the rgb space component and cut apart, obtain more accurately target insulator zone, by the weights of Genetic Algorithm Optimized Neural Network, its result more accurately effectively;

2) can under the situation that does not have a power failure, does not take a sample, do not disintegrate, detect by the filthy state to insulator, also need any equipment be installed on shaft tower, the operation and maintenance cost is low, simple to operate, safe and reliable, and accuracy of detection is high.

The accompanying drawing explanation

Fig. 1 is process flow diagram of the present invention;

The mean value of Fig. 2 target that is 20 width samples in gray-scale map and G divide spirogram and background poor;

The mean value of Fig. 3 target that is 20 width samples in gray-scale map and R divide spirogram poor.

Embodiment

Below in conjunction with accompanying drawing, the specific embodiment of the present invention is described further.

As shown in Figure 1, the invention provides a kind of transmission line of alternation current based on infrared thermal imagery insulator contamination condition detection method, comprise the following steps:

Step 1, take Insulator Infrared Image, the requirement insulator is clear and legible, shooting background is aimed at sky as far as possible, because the sky background temperature of environment temperature and infrared image shooting is inconsistent, need to take the environment temperature of obtaining simultaneously while taking, and, because of contaminated insulator fever phenomenon difference under different humidity, need to measure the ambient humidity while taking.

Step 2, adopt neighborhood averaging to carry out denoising to image, establishes original image and be

, in image, get respectively a neighborhood S centered by each pixel, calculate the weighted mean value of all pixel grey scales in S, as the pixel value of image after spatial domain smoothing processing.That is:

In formula, M is the pixel number in neighborhood S, and by contrasting the smooth effect of each radius of neighbourhood, the present invention adopts

Neighborhood averaging.

Step 3, extract R component and the G component of thermography rgb space, the basis of cutting apart as follow-up insulator.The red R component of infrared image and green G component can raise along with the rising of temperature, compare directly and process with gray-scale map, and the discrimination of background and target is larger.As shown in Figure 2,

From Fig. 2 and Fig. 3, finding out, in the G component, the difference of the mean value of target and background, apparently higher than gray-scale map, is easier to insulator and cuts apart, and in the R component, the mean value of target maintains more than 250 substantially, can be used as auxiliary component and carrys out cutting apart of accurate insulator.

Step 4, adopt the seed region growth method to R, to divide spirogram and G to divide spirogram to cut apart respectively.The present invention adds up acquisition RG by sample image and divides spirogram seed region selected threshold and growth conditions threshold value, order IMean the view picture infrared image,

, Be respectively image IRed color component value and green component values, be infrared image ISeed region, the red component threshold value is

, the green component threshold value is

, the seed region expression formula is

Red component figure seed growth condition is as follows:

1) find out the seed region edge, as initial seed point;

2) establishing Seed Points is

, the judgement Seed Points

Whether the pixel in neighborhood meets the following conditions:

If meet, by point

Be considered as new Seed Points.Wherein,

For the threshold value of red component figure Central Plains Seed Points with the difference of the Seed Points of new growth,

For insulator region threshold in red component figure;

3), to new Seed Points repeating step (2), until no longer include pixel, satisfy condition and stop growing, thereby obtain final red component seed region I _SR .

In like manner, according to the different threshold values of statistics, obtain the green component seed region I _SG .

Step 5, the R after cutting apart divides spirogram and G to divide spirogram to merge, and obtains final insulator and cuts apart image H:

Step 6, extract the insulator contamination feature, and characteristic quantity is: the maximum temperature in insulator chain card zone, minimum temperature, medial temperature, Temperature Distribution variance, three of the Temperature Distribution conic fitting parameters on the insulator axis.The data based insulator of characteristic quantity is cut apart image and infrared temperature data acquisition.

Step 7, detect insulator dirty degree based on the GA-BP neural network.Floor operation Insulator Infrared Image is as sample graph, the gradation for surface pollution of sample graph is measured by equivalent salt density, using environment temperature, humidity and seven characteristic quantities as input quantity, five gradation for surface pollution are as output quantity, set up a BP neural network, and utilize genetic algorithm to be optimized the network node weights.

The mathematical description of BP neural network algorithm is as follows:

Wherein

For the total error of network,

For desired output,

For the actual output of network.

Adopt genetic algorithm to be optimized step to the BP neural network as follows:

1) set up the BP neural network, determine output input node number X, Y, the implicit number of plies H, frequency of training N, training error

, it is 9 dimensions that the present invention inputs the node number, output node is 5 dimensions, and corresponding five gradation for surface pollution, the implicit number of plies is set to 2 layers, training iteration 2000 times, error is set and is less than 0.05%;

2) produce initial population, one of initialization comprises n chromosomal initial population

, and, to its coding, determine maximum iteration time T, crossover probability P _c With the variation probability P _m , BP network weight and threshold value are cascaded up in certain sequence, as a chromosome of genetic algorithm, its length l is weights number and the threshold value number sum of neural network, namely

3) fitness function calculates, by the global error function of BP network As fitness function, calculate each chromosomal fitness in population, and descending sort by size;

4) genetic evolution, act on the parent population by the crossover and mutation operator P _t Produce progeny population Q _t , and by the population of two population gangs formation 2n R _t , to having the population of 2n scale R _t Carry out sort operation, according to the principle generation population of future generation of crowded selection operator P _{t+ 1} , and copy replication is arrived to P'.

5) repeat (3) (4) step, until reach maximum genetic algebra, by optimum individual decoding in P', as initial weight and the threshold value of BP network, then according to BP Learning Algorithm training network.

Last installation procedure on computers, the infrared image be written into temperature data can obtain the gradation for surface pollution that this insulator is corresponding.

Claims (1)

1. the insulator contamination condition detection method of the transmission line of alternation current based on infrared thermal imagery, is characterized in that, comprises the following steps:

(1) Infrared Thermogram of detected insulator picked-up;

(2) adopt neighborhood averaging to carry out the noise elimination to colored thermography;

(3) extract red R component and the green G component of thermography rgb space, compared to gray-scale map, G divides in spirogram the discrimination of insulator and background larger, and R divides insulator in spirogram to maintain higher pixel value, further auxiliary partition;

(4) adopt the seed region growth method to R, to divide spirogram and G to divide spirogram to cut apart respectively;

(5) R after cutting apart divides spirogram and G to divide spirogram to merge, and obtains final insulator and cuts apart image;

(6) extract the insulator contamination feature, characteristic quantity is: the maximum temperature in insulator chain card zone, minimum temperature, medial temperature, Temperature Distribution variance, three of the Temperature Distribution conic fitting parameters on the insulator axis;

(7) based on the GA-BP neural network, detect insulator dirty degree, take Insulator Infrared Image in service as sample graph, using environment temperature, humidity and seven characteristic quantities as input quantity, five gradation for surface pollution are as output quantity, set up a BP neural network, and utilizing genetic algorithm to be optimized the network node weights, the infrared image be written into temperature data namely obtains the gradation for surface pollution that this insulator is corresponding.

Images